> p.s. David Luyer added:
> > Although, if you look at the old SPEC benchmarks for Alphas, you'll
> > find
> > a 4 line long list of compiler options, among them options to use 32
> > bit
> > longs and pointers. While the Alpha is a 64-bit machine, it must have
> > benchmarked better as 32-bit initially.
>
> Another possibilty, which I would be inclined strongly suspect, is that
> test suite was broken and whomever was running tests took an easy way
> out.
*bzzzzzzt* wrong.
The "standard" (compiler/portability) flags don't use xtaso, so it is 64-bit
clean code.
The "peak" flags use xtaso_short - it's an optimization.
This is in the specbench.org cpu95 results submitted by DEC. They also
used the magic -32data flag for some programs.
elysium; cat test.c
main() {
printf("%d %d\n", sizeof(char *), sizeof(long));
}
elysium; cc -xtaso_short -32data test.c -o test
elysium; ./test
4 4
elysium;
BTW: That answers someone else's answer about 32 bit longs. It's just
undocumented.
David.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/